Skip to content

docs: record the audit backlog, replacing the lost tracker database - #14

Merged
Ugbot merged 1 commit into
mainfrom
docs-audit-backlog
Jul 27, 2026
Merged

docs: record the audit backlog, replacing the lost tracker database#14
Ugbot merged 1 commit into
mainfrom
docs-audit-backlog

Conversation

@Ugbot

@Ugbot Ugbot commented Jul 27, 2026

Copy link
Copy Markdown
Owner

The 2026-07-25 audit produced 44 tickets in a SQLite tracker under the home directory. That database was destroyed on 2026-07-26 by an unrelated incident that wiped most of ~. The eleven epics survived as issues #1#11; the thirty-three sub-tasks existed only in the database.

docs/audit-backlog.md is now the source of truth for them — committed, so it survives any future machine loss.

What it preserves

  • All 44 tickets, each with the original file:line citations and the measured numbers, so nothing has to be rediscovered.
  • The epic → issue map with each task's tracker uid, recovered from the markers in the issue bodies. A rebuilt tracker can reuse those uids, so the existing issue markers keep resolving instead of renumbering.
  • The measured baseline (770 main-module tests, 922 in CI, per-surface counts) and the four ground-rule decisions that shaped the list — ship scope, the paradigm bar, the publishing target, and Java 21.
  • Two tickets already closed, marked as such: AGS-9 (f39e6a1) and AGS-22 (9731fdc).

Corrections to the audit, recorded deliberately

Six findings were wrong on first pass and are documented as such so they are not re-filed as defects:

  • The langchain4j and a2a-java BOMs resolve fine — the offline failure was the unpublished jagentic-core.
  • The shaded-jar setup is correct (thin main artifact + classified -uber), not an anti-pattern.
  • examples-bin/ is intentionally tracked scripts, not committed build output.
  • A mistyped removeEldestEntry does not compile as a silent overload — javac rejects the erasure clash, so the compiler guards that bug, not a test.
  • tasks.uid already existed in the tracker with deterministic re-keying on collision; the real defect was publishing the mutable key in issue markers.
  • ToolExecutorRegistry.register(...) does exist; only ToolRegistry lacks it.

Verification

Docs-only change, no code touched. Rebuilt on a freshly restored toolchain (JDK 21.0.12 + Maven 3.9.16 via sdkman, from a completely empty ~/.m2): jagentic-core 95 tests pass, main module rebuild green. CI will confirm on this PR.

🤖 Generated with Claude Code

The 2026-07-25 audit produced 44 tickets in a SQLite tracker under the home
directory. That database was destroyed on 2026-07-26 by an unrelated incident
that wiped most of ~. The eleven epics survived as GitHub issues #1-#11; the
thirty-three sub-tasks existed only in the database.

This file is now the source of truth for them. Each entry keeps the file:line
citations and the measured numbers from the audit, so nothing has to be
rediscovered. It also records the epic -> issue map with each task's tracker
uid, recovered from the marker in the issue bodies, so a rebuilt tracker can
reuse the same uids and keep the existing issue markers resolving.

Includes a "corrections to the audit itself" section: six findings that were
wrong on first pass (the BOMs resolve fine; the shaded-jar split is correct;
examples-bin is not build output; a mistyped removeEldestEntry does not compile
as a silent overload; tasks.uid already existed with deterministic re-keying;
ToolExecutorRegistry.register does exist). Recorded so they are not re-filed as
defects.

Two tickets are already closed and marked as such: AGS-9 (f39e6a1) and
AGS-22 (9731fdc).

Co-Authored-By: Claude <noreply@anthropic.com>
@Ugbot
Ugbot merged commit 9157cae into main Jul 27, 2026
4 checks passed
@Ugbot
Ugbot deleted the docs-audit-backlog branch July 27, 2026 00:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant